Random forests for time-dependent processes
Abstract
Random forests were introduced by Breiman in 2001. We are interested in the theoretical study of both the random forest-random input and a simplified version of the random forest: the centred random forest. Let ((X_1 , Y_1) ,. .. , (X_n , Y_n)) be random variables. Under the independent and identically distributed hypothesis, Biau studied the simplified version and got rate of convergence in the sparse case. Biau, Scornet and Vert proved the consistency of the original algorithm when the regression model follows an additive model and that X~Unif (0, 1)^p. However we are commonly faced to applications where the i.i.d hypothesis is not satisfied for example when dealing with time series. We extend the previous results to the case where observations are weakly dependent. .
Fichier principal
Random_forests_for_time_dependent_processes.pdf (173.95 Ko)
Télécharger le fichier
Origin : Files produced by the author(s)
Loading...